MATH 110: LINEAR ALGEBRA

FALL 2007/08
PROBLEM SET 7 SOLUTIONS

Let V be a vector space. The identity transformation on V is denoted by IV , ie. IV : V → V and

IV (u) = u for all u ∈ V . The zero transformation on V is denoted by OV , ie. OV : V → V and
OV (u) = 0V for all u ∈ V where 0V is the zero vector/additive identity of V .

1. Find the kernels of the following linear transformation.

(a) Ta : R2 → R2 defined by
Ta (x, y) = (2x + 3y, 7x − 5y)
for all (x, y) ∈ R2 .
Solution. Ta (x, y) = (0, 0) iff (2x + 3y, 7x − 5y) = (0, 0) iff x = y = 0. Hence ker(Ta ) =
{(0, 0)}.
(b) Tb : R2 → R3 defined by
Tb (x, y) = (x, 0, 0)
for all (x, y) ∈ R2 .
Solution. Tb (x, y) = (0, 0, 0) iff x = 0. Hence ker(Tb ) = {(0, y) ∈ R2 | y ∈ R} =
span{(0, 1)}.
(c) Tc : R2 → R defined by
Tc (x, y) = x + 2y
for all (x, y) ∈ R2 .
Solution. Tc (x, y) = 0 iff x = −2y. Hence ker(Tc ) = {(−2y, y) ∈ R2 | y ∈ R} =
span{(−2, 1)}.
(d) Td : P5 → P10 defined by
Td (p(x)) = p(x2 )
for all p(x) ∈ P5 .
Solution. Let 0n = 0 + 0x + · · · + 0xn . Td (p(x)) = 010 iff p(x2 ) = 010 iff p(x) = 05 .
Hence ker(Td ) = {05 }.
(e) Te : P5 → P4 defined by
Te (p(x)) = p0 (x)
for all p(x) ∈ P5 .
Solution. Te (p(x)) = 04 iff p0 (x) = 04 iff p(x) = c (constant polynomial). Hence
ker(Te ) = {c} = span{1}.
(f) Tf : P5 → P6 defined by
Z x+9
Tf (p(x)) = p(t) dt
−3
for all p(x) ∈ P5 .
R x+9
Solution. Tf (p(x)) = 06 iff −3 p(t) dt = 06 iff p(x) = 05 . Hence ker(Tf ) = {05 }.

2. Let T : U → V and S : V → W be linear transformations.

Date: November 4, 2007 (Version 1.0).

1
 (a) Show that
im(S ◦ T ) ⊆ im(S)
and
ker(S ◦ T ) ⊇ ker(T ).
Solution. Let w ∈ im(S ◦ T ). Then there exist some u ∈ U such that w = S ◦ T (u) =
S(T (u)). So w ∈ im(S) since w = S(v) where v = T (u) ∈ V . Hence im(S ◦ T ) ⊆
im(S). For the second part, let u ∈ ker(T ). Then T (u) = 0V and so S(T (u)) = S(0V ),
ie. S ◦ T (u) = 0W , ie. u ∈ ker(S ◦ T ). Hence ker(T ) ⊆ ker(S ◦ T ).
(b) Show that if T is surjective, then
im(S ◦ T ) = im(S).
Solution. By part (a), we know that im(S ◦ T ) ⊆ im(S). We will show the reverse
inclusion. If w ∈ im(S), then there exists some v ∈ V such that w = S(v). But since
T : U → V is surjective, there exists some u ∈ U such that v = T (u). So w = S(T (u)) =
S ◦ T (u), ie. w ∈ im(S ◦ T ). Hence im(S) ⊆ im(S ◦ T ).
(c) Show that if S is injective, then
ker(S ◦ T ) = ker(T ).
Solution. By part (a), we know that ker(S ◦ T ) ⊇ ker(T ). We will show the reverse
inclusion. If u ∈ ker(S ◦ T ), then S(T (u)) = S ◦ T (u) = 0W . But since S is injective
(and linear), by a result in the lectures, we ker(S) = {0V }. So T (u) ∈ ker(S) implies that
T (u) = 0V , which in turns implies that u ∈ ker(T ). Hence ker(S ◦ T ) ⊆ ker(T ).

3. Let V be a finite-dimensional vector space. Let S : V → V and T : V → V be linear operators

on V .
(a) Show that
rank(S ◦ T ) ≤ min{rank(S), rank(T )}
and
nullity(S ◦ T ) ≤ nullity(S) + nullity(T ).
Solution. By Homework 6, Problem 3(b), we have that
rank(S ◦ T ) ≤ rank(S) and rank(S ◦ T ) ≤ rank(T ).
Therefore
rank(S ◦ T ) ≤ min{rank(S), rank(T )}.
For the second part, consider the function ϕ : ker(S ◦ T ) → V defined by ϕ(v) = T (v)
for all v ∈ ker(S ◦ T ) (note that ϕ and T are two different functions because they have
different domains). Since ϕ(αu + βv) = T (αu + βv) = αT (u) + βT (v) = αϕ(u) + βϕ(v),
the function ϕ is a linear transformation from the vector space ker(S ◦ T ) to the vector
space V . Now applying the rank-nullity theorem in the lectures to ϕ, we get
dim(ker(S ◦ T )) = nullity(ϕ) + rank(ϕ) = dim(ker(ϕ)) + dim(im(ϕ)). (3.1)
If w ∈ im(ϕ), then w = ϕ(v) for some v ∈ ker(S ◦ T ) and
S(w) = S(ϕ(v)) = S(T (v)) = S ◦ T (v) = 0
and so w ∈ ker(S). Hence im(ϕ) ⊆ ker(S) and so
dim(im(ϕ)) ≤ dim(ker(S)) = nullity(S). (3.2)
If v ∈ ker(ϕ), then
0 = ϕ(v) = T (v)
and so v ∈ ker(T ). Hence ker(ϕ) ⊆ ker(T ) and so
dim(ker(ϕ)) ≤ dim(ker(T )) = nullity(T ). (3.3)

2
 Applying (3.2) and (3.3) to the rhs of (3.1), we obtain
nullity(S ◦ T ) ≤ nullity(S) + nullity(T ).
(b) Show that
rank(S + T ) ≤ rank(S) + rank(T ).
Solution. Let w ∈ im(S + T ). Then w = (S + T )(v) for some v ∈ V . Observe that
w = (S + T )(v) = S(v) + T (v) ∈ im(S) + im(T ).
Hence im(S + T ) ⊆ im(S) + im(T ) and hence
dim(im(S + T )) ≤ dim(im(S) + im(T )) ≤ dim(im(S)) + dim(im(T )),
where the second inequality follows from Homework 4, Problem 4(a). In other words, we
have
rank(S + T ) ≤ rank(S) + rank(T ).
(c) Show that if S ◦ T = OV , then
rank(S) + rank(T ) ≤ dim(V ).
Solution. Note that ker(OV ) = V — for any v ∈ V , OV (v) = 0, so V ⊆ ker(OV ); on
the other hand, we always have ker(OV ) ⊆ V . Hence nullity(OV ) = dim(V ). Now we will
apply part (a),
nullity(S ◦ T ) ≤ nullity(S) + nullity(T )
to get
dim(V ) ≤ nullity(S) + nullity(T ).
Adding dim(V ) to both sides of the inequality and bringing the two terms on the rhs to
the lhs, we get
dim(V ) − nullity(S) + dim(V ) − nullity(T ) ≤ dim(V ).
Finally, we apply the rank-nullity theorem twice to get
rank(S) + rank(T ) ≤ dim(V ).

4. Let V be a finite-dimensional vector space. Let T : V → V be a linear operator on V . Show

that the following three statements are equivalent:
(i) V = ker(T ) + im(T );
(ii) ker(T ) ∩ im(T ) = {0V };
(iii) V = ker(T ) ⊕ im(T ).
Solution. By Homework 4, Problem 4(a),
dim(ker(T ) + im(T )) + dim(ker(T ) ∩ im(T )) = dim(ker(T )) + dim(im(T ))
= nullity(T ) + rank(T )
= dim(V ).
(a) ⇒ (b): if ker(T ) + im(T ) = V , then we get
dim(V ) + dim(ker(T ) ∩ im(T )) = dim(V ),
ie.
dim(ker(T ) ∩ im(T )) = 0,
ie. ker(T ) ∩ im(T ) = {0V }.
(b) ⇒ (c): if ker(T ) ∩ im(T ) = {0V }, then we get
dim(ker(T ) + im(T )) + dim({0V }) = dim(V )
ie.
dim(ker(T ) + im(T )) = dim(V ).

3
 Since ker(T ) + im(T ) is a subspace of V , the only way their dimensions could be equal is if
ker(T ) + im(T ) = V.
But since ker(T ) ∩ im(T ) = {0V }, the sum is a direct sum, ie.
ker(T ) ⊕ im(T ) = V.
(c) ⇒ (a): ker(T ) ⊕ im(T ) = V implies both (a) and (b), so it implies (a) in particular.

5. Let T : R3 → R3 be defined by T (x, y, z) = (2x, 4x − y, 2x + 3y − z).

(a) Show that T is invertible.
Solution. Let (x, y, z) ∈ ker(T ). Then T (x, y, z) = (2x, 4x − y, 2x + 3y − z) = (0, 0, 0).
Solving this, we get x = y = z = 0 and so (x, y, z) = (0, 0, 0) and so ker(T ) = {(0, 0, 0)}.
Hence T is invertible by a result in the lectures.
(b) Find a formula for T −1 .
Solution. Set T (x, y, z) = (r, s, t) and so T −1 (r, s, t) = (x, y, z). Now to find a formula
for T −1 we solve (2x, 4x − y, 2x + 3y − z) = (r, s, t) for x, y, z to get (x, y, z) = ( 12 r, 2r −
s, 7r − 3s − t). Hence T −1 (r, s, t) = ( 12 r, 2r − s, 7r − 3s − t).
(c) Find a formula for T −2 := T −1 ◦ T −1 .
Solution. T −2 (r, s, t) = T −1 ( 12 r, 2r − s, 7r − 3s − t) = ( 14 r, −r + s, − 19
2 r + 6s + t).
(d) Let B = {e1 , e2 , e3 } be the standard basis for R3 . Find the coordinate representations
A = [T ]B,B ∈ R3×3 and B = [T −1 ]B,B ∈ R3×3
of T and T −1 with respect to the basis B on both the domain and the codomain.
Solution. Since T (1, 0, 0) = (2, 4, 2), T (0, 1, 0) = (0, −1, 3), T (0, 0, 1) = (0, 0, −1), we
get  
2 0 0
A = [T ]B,B = 4 −1 0  .
2 3 −1
Likewise, since T −1 (1, 0, 0) = ( 21 , 2, 7), T −1 (0, 1, 0) = (0, −1, −3), T (0, 0, 1) = (0, 0, −1), we
get 1 
2 0 0
B = [T −1 ]B,B =  2 −1 0  .
7 −3 −1
(e) Are the matrices A and B nonsingular (a.k.a. invertible)? If so, what are A−1 and B −1 ?
Solution. Note that AB = I = BA. So A and B are both nonsingular and A−1 = B,
B −1 = A.